Infineon advances GaN power conversion for 800 V DC AI data centres
Infineon Technologies is expanding its collaboration with NVIDIA on 800 V direct-current (DC) power architectures for AI data centres, focusing on gallium nitride (GaN) technology for high-voltage intermediate bus converters (IBCs).
The announcement, timed to coincide with NVIDIA’s GTC Berlin developer conference, builds on joint work announced in May 2025 and further detailed in October 2025. That work included power architectures designed to support serviceable, hot-swappable server racks operating from an 800 V DC distribution system.
The latest collaboration focuses on the IBC stage, which converts the high-voltage DC bus into a lower voltage for downstream power-conversion stages supplying AI processors and other server components.
“AI is fundamentally reshaping the demands we place on power infrastructure, and semiconductors are at the heart of solving that challenge,” said Jochen Hanebeck, chief executive officer of Infineon Technologies. “Our collaboration with NVIDIA on 800 Volt power architectures shows what is possible when deep semiconductor expertise meets the pace of innovation required by AI. Gallium nitride-based system solutions are a key enabler of this next generation of data centres, and Infineon is proud to help power the AI era from grid to core.”
GaN targets high-density AI data centre power conversion
As AI workloads increase server power consumption, data centre designers are evaluating higher-voltage DC distribution to support greater power delivery while managing losses, thermal constraints and limited space.
In an intermediate-bus architecture, the IBC steps down the main DC bus voltage to a lower-voltage rail. Further conversion stages then generate the low-voltage supplies required by processors and other components.
An 800 V DC architecture is emerging as an additional option alongside existing AC and lower-voltage DC distribution approaches. Its adoption is intended to address the power-delivery challenges associated with increasingly dense AI computing systems.
GaN power semiconductors are well suited to high-frequency power conversion. Compared with conventional silicon devices, GaN can support faster switching and lower switching losses in appropriately designed circuits.
For IBC applications, switching frequencies of up to 1 MHz can enable smaller magnetic components and passive components. This can help increase power density and reduce converter height, valuable characteristics in densely packed AI server racks.
Lower conduction and switching losses can also improve conversion efficiency and reduce heat generation. These benefits become increasingly important as rack power requirements rise towards the megawatt range.
Infineon is combining its GaN portfolio with PSOC Control C3 microcontrollers and EiceDRIVER gate-driver products to provide system-level solutions for IBC designs.
Conversion options for 800 V DC power systems
Infineon’s GaN portfolio supports several voltage-conversion ratios for high-voltage IBC designs, including:
800 V to 50 V
800 V to 12 V
800 V to 6 V
These options give designers flexibility when developing rack-level and server-level power architectures. The choice of conversion ratio depends on the wider system design, including downstream conversion stages, power requirements and efficiency targets.
Infineon’s power semiconductor portfolio spans silicon, GaN and silicon carbide (SiC) technologies, allowing the company to address different stages and requirements across the data centre power chain. Its capabilities include in-house semiconductor manufacturing, as well as application and system-level expertise.
Collaboration extends across the data centre power chain
The Infineon–NVIDIA collaboration also covers power-path protection. Work announced in October 2025 addressed safety and servicing requirements for 800 V DC systems, including hot-swap controller functionality intended to support the replacement or servicing of server components.
Together, the companies’ work covers multiple stages of data centre power delivery, from grid-facing power supply units through intermediate conversion and distribution to the point of load at the processor.
As AI infrastructure continues to scale, high-efficiency power conversion, power density and serviceability are becoming increasingly important design considerations for next-generation data centres.
